yearbook2016-cut

지도교수 Study
학위논문
김민기 졸업연구

박사과정
고건혁

A Bayesian Learning Model for Measuring Social In-
fluence: New Music Diffusion on Twitter

As online sharing of music through social media emerges as a promising mode for
diffusion of new music, firms are now interested in identifying individuals who have
strong influence on other listeners. However, previous studies on disproportionate
social influence have neglected the fact that in real-world social networks, indi-
viduals not only occupy certain social positions, but also have heterogeneous social
relationships with others. Exploiting coexistence of asymmetric and symmetric rela-
tions in the microblogging service Twitter, this study simultaneously examines the
effects of network structure such as indegree centrality and dyadic properties of
interpersonal relationships, including communication frequency and net-work em-
beddedness, on the diffusion of new music. To the best of my knowledge, this is the
first empirical research that proposes a new viral marketing strategy that reflects
the concur-rent impact of social position and of dyadic relation characteristics on
disproportionate peer influence. I develop a model for an individual’s information
sharing decision where each Twitter user updates her/his belief about the quality of
musicians in a Bayesian learning fashion. Multiple information sources for learning
are classified into eight different types of online peers. Utilizing Twitter users’ actual
behavior data related to new musicians of the audition reality show, I provide an
empirical demonstration of estimating each peer type’s social influence and practical
managerial insights based on estimated results. Results of a what-if scenario analy-
sis show that the strategy deploying the proposed model outperforms traditional
word-of-mouth (WOM) marketing strategies, including hub targeting and random
seeding, in its return on investment (ROI).

51

Study 지도교수
학위논문
졸업연구 이지현

박사과정
김선중

A Study on Metadata Structure and Recommenders
of Biological Systems to Support Bio-inspired De-
sign. Graduate School of Culture Technology

Bio-inspired design was introduced as an alternative method to encourage break-
through innovations during design projects by stimulating analogical reasoning and
thinking of designers. However, the method did not perform as well as research-
ers expected because most designers, who are novices in the fields of biology and
ecology, cannot infer the proper analogue (i.e. biological system) from nature. To
resolve this fundamental problem, a causal model based representation framework
for ‘analogical reasoning’ – searching and selecting the biological systems to apply –
have been developed. In addition, ontology based repository structures and retrieval
systems have been proposed to support ‘analogical thinking’ of designers. Never-
theless, these systematic approaches still restrict the candidates and inevitably lose
potential biological systems relevant to the design project, due to the ‘physical rela-
tion’ biased problem and the ambiguity of the indexing mechanism of both current
representation frameworks and retrieval systems. For example, the causality based
support system known as a robust representation framework for a single biological
system, stores information of a biological system only by its internal ‘physical rela-
tions’ and retrieves biological systems only by the physical relevance. However,
from the perspective of ecological thinking, the further relatedness of ‘physical, bio-
logical, and ecological relations’ composes the holistic concept used to identify an
organism in the flow of evolution because the ‘biological and ecological relations’
are also involved in the traits that designers may be interested in. Therefore, the
supplementary information for ‘biological and ecological relations’ must be added
to index the biological and environmental interactions, and to use the connectivity
among entire organisms in the retrieval process. In this research, a causality based
holistic representation framework for biological systems and an ‘all-connected’ on-
tology based repository and retrieval system are developed to support bio-inspired
design. This effort might provide more opportunities in a bio-inspired design pro-
cess by adding potential biological systems that might previously not have been
considered.

52

지도교수 Study
학위논문
남주한 졸업연구

박사과정
김승훈

Sonic Control of Audio Feedback and its Artistic
Uses according to Environmental Openness

Audio feedback, in which sound is generated through the connection of a micro-
phone, signal processing components and a loudspeaker, is one of the popular struc-
tures used in computer music. Since audio feedback has a circular structure using
the output signals as re-input signal, beyond the one-way relationship, it results in
unique features and a lot of attempts have been made to take advantage of an audio
feedback in music composition and sound arts.

This dissertation practically explores audio feedback systems by developing con-
trol methods and applications according to environmental openness. Environmental
openness, degree of contact with external energy and information, enables develop-
ment of various musical applications depending on the structure. Since audio feed-
back features nonlinear characteristics, sonic control methods are required to create
the feedback-based applications for specific purposes (e.g., performance, composi-
tion and exhibition). These are the reasons why this thesis includes control methods
as well as artistic uses. By generating sounds using the feedback structure according
to environmental openness and presenting control methods for certain sonic prop-
erties, we propose creative utilization of feedback systems in the fields of musical
interactions while preserving the unique nonlinear characteristics. Classification of
these methods also builds a new framework of audio feedback systems regarding
sonic control, which would explore a wide range of potential applications.

We present several systems in three types of environmental openness: closed, semi-
open and open. Closed feedback structure is a typical model of Karplus-Strong algo-
rithm and digital waveguide synthesis, and we propose synthesis models and instru-
ment software for the Geomungo, a Korean traditional plucked-string instrument,
with a time-varying filter for vibrato effects. Semi-open feedback system installed
in a specific shape space (e.g., a pipe) generates resonant sounds, and we propose
two types of feedback-based pipe interfaces. Open system could generate various
timbres depending on the internal and external conditions, and we explore the meth-
ods for intentional control of audio feedback through a context-based control and an
evolutionary control.

53

Study 지도교수
학위논문
졸업연구 차미영

박사과정
박자람

Analysis of Cultural Values in Online Communica-
tion across Countries on the Use of Nonverbal Cues
and Content Consumption

From the ancient time of, Aristotle and Confucius, people across different cultures
have perceived the world in different ways (Nisbett, 2010). Culture influences every
aspect of our lives from basic processes of cognition and perception (Ji et al., 2000;
Masuda and Nisbett, 2001) to a range of our behavior (Brislin, 2006). Therefore,
studying cultural differences and cultural distance is an important key to under-
stand human behavior in the analysis of social science or humanities (Baldwin et
al., 2006), and the concepts of cultural distance (e.g., Hofstede’s cultural dimension
theory and psychic distance) are widely used for quantifying cultural differences.
Recently, the advent of Web has fundamentally changed the way individuals are ex-
posed to other cultures. Web gives us fruitful opportunities to consume the cultural
content from other culture; we are now able to communicate with people across
the globe easier than before. However, online communication (e.g, user interaction,
online contents co-consumption) is still affected by national-level factors such as
language (Ronen et al., 2014), geographical border (Kulshrestha et al., 2012; Park
et al., 2015; Platt et al., 2015), economic status (Barnett and Sung, 2005; Platt et
al., 2015), religion (Park et al., 2015), and cultural values (Garc ́ıa- Gavilanes et al.,
2013, 2014).

Such phenomenon motivates this dissertation: my thesis focuses on 1) the impact of
cultural difference on online communication in order to provide insights and impli-
cations for refining and advancing cross-cultural theories, and 2) a framework for
measuring and finding important culture values influencing online communication
across cultures. Thanks to the recorded web data containing most of all daily life of
people, we are able to analyze large-scale human behavior data naturally generated
from all around world.

In this dissertation, I first propose a methodology for analyzing the impact of culture
on online communication using the large-scale online dataset. Second, I character-
ize cultural differences in online communication by focusing on the nonverbal cues
in microblogs, Twitter, and video consumption patterns on YouTube. The results
demonstrate that the use of representative nonverbal cues, emoticons, and the video
co-consumption patterns are related to users’ cultural background. Finally, I design a
model for finding the main cultural values related to online communication by rely-
ing on Gudykunst’s cultural variability in communication theory, and then figure out
the relationship between cultural values and online communication.

54

지도교수 Study
학위논문
원광연 졸업연구

박사과정
이상원

Verification of the Spatial Design Adequacy for the
Prospective Occupants’ Amenity and Efficiency
Needs

The best spatial design condition to satisfy the occupancy needs of amenity and ef-
ficiency is determined through analyzing the spatial design adequacy (SDA). In this
study, the relationship between the space design elements and space on future occu-
pants’ perception are analyzed. A series of two experiments were conducted in order
to verify the six hypotheses. The findings successfully support the all hypotheses
of this study. The findings of experiments demonstrate that the causal relationship
between the spatial factors and space is clearly present in the occupants’ perception,
reflecting the time-sequential characteristics of the actual experience divided into
amenity and efficiency. This result indicates that the correlation between the spatial
factors and space of SDA under the occupants’ perception processing elaboration
can be a useful guide to predict the occupancy satisfaction of amenity and efficiency
in real spaces.

55

Study 지도교수
학위논문
졸업연구 시정곤

박사과정
이지현

The Framework of Systematic Creation Process and
Methods for Designing Novel Conceptual Charac-
ter’s Appearance

Novel conceptual characters appear in games, movies, and other virtual media related to
myths, legends, and future worlds. Newly created characters have expanded functionality
combining parts of other creatures, artefacts, science, and technology; and using cross-
ontological synthesizing methods either consciously or unconsciously. The aim of this
study is to propose a systematic process and specific methods that can be used to create
novel conceptual characters. Therefore, to find the specific steps and methods required
for each stage on the basis of the design process, including analysis, synthesis, and evalu-
ation, each stage was systemized by analyzing related studies and conducting case stud-
ies, including character design implementation.
First, the study in the analysis stage proposes a structure for extracting a character’s ap-
pearance components from a narrative for representing character’s appearance. Second,
the study in the synthesis step is performed to propose synthetic character generation
strategies that can be used to create novel conceptual characters through both the analysis
of virtual characters that have appeared in games and movies and the application of con-
cept combination strategies used in cognitive psychology and cognitive science. Third,
the study in the evaluation stage concerns the evaluation criteria for novel conceptual
characters. The evaluation criteria of emergence, practicality, and originality were es-
tablished based on creativity studies of novel concept combination and creative design
output. In addition, the effectiveness of the structure and methods produced through each
step was verified by a comparative t-test analysis of the case study results to design novel
conceptual characters.
The establishment of a systematic and specific process and methods for designing char-
acters have been difficult to organize academically. Therefore, the creation process and
methods for designing novel conceptual character’s appearance produced in this study
have the meaning that establishes a systematic creation process and evaluation methods
for designing all types of characters.

56

지도교수 Study
학위논문
노준용 졸업연구

박사과정
최병국

Sketch-based Motion Editing for 3D Articulated
Characters

In this doctoral dissertation, a novel approach for the expressive editing of 3D artic-
ulated character motion is presented. The proposed approach solves for the motion
given a set of projective constraints that relate the sketch inputs to the unknown 3D
poses. This dissertation introduces the concept of sketch space, a contextual geomet-
ric representation of sketch targets ―motion properties that are editable via sketch
input― that enhances, right on the viewport, different aspects of the motion. The
combination of the proposed sketch targets and space allows for seamless editing of
a wide range of properties, from simple joint trajectories to local parent-child spati-
otemporal relationships and more abstract properties such as coordinated motions.
This is made possible by interpreting the user’s input through a new sketch-based
optimization engine in a uniform way. In addition, the proposed view-dependent
sketch space also serves the purpose of disambiguating the user inputs by visual-
izing their range of effect and transparently defining the necessary constraints to set
the temporal boundaries for the optimization.

57

Study 지도교수
학위논문
졸업연구 이지현

박사과정
현경훈

Augmenting gestural strokes with dynamics-driven
simulations for immersive freeform 3d modeling

This research proposes objective measures to evaluate product designs through
style quantification. To do that, five tasks were conducted: first, yearly car mod-
els were collected. Second, design elements were extracted. Third, a style analysis
was conducted through similarity calculation and visual significance identification.
Fourth, styling strategy positioning of the yearly car design models was analyzed.
Lastly, a genetic algorithm was used to synthesize the car designs that maintain
brand’s unique style. From these five tasks, the styling strategy can be quantitatively
analyzed and be implemented to automatically synthesize optimized design alterna-
tives. In addition, research results and in-depth interviews with professional design-
ers from diverse nationalities confirmed the validity of this research. The results of
the research open up new possibilities for making more effective strategic design
decisions for managerial decisions through quantitative analysis of styling and pro-
vide new insights that were not possible to discover before. Based on this research,
researchers and designers can empirically analyze the influence of styling on car de-
sign industries, the effect of the family look within car brands, visual significance of
design elements and the design trend transitions to grope for a more effective design
management.

58

2016 ANNUAL
REPORT
Graduate School of Culture Technology



Project

학술 활동
창작 활동

Project
학술활동

국제저널

An Eulerian approach for constructing a map between sur-
faces with different topologies

Computer Graphics Forum
Hangil Park, Youngjin Cho, Seungbae Bang and Sung-Hee Lee

3D objects of the same kind often have different topologies, and finding correspond-
ence between them is important for operations such as morphing, attribute transfer,
and shape matching. This paper presents a novel method to find the surface corre-
spondence between topologically different surfaces. The method is characterized by
deforming the source polygonal mesh to match the target mesh by using the inter-
mediate implicit surfaces, and by performing a topological surgery at the appropriate
locations on the mesh. In particular, we propose a mathematically well-defined way
to detect the topology change of surface by finding the non-degenerate saddle points
of the velocity fields that tracks implicit surfaces. We show the effectiveness and
possible applications of the proposed method through several experiments.

62

Project
학술활동

국제저널

Augmenting Environmental Interaction in Audio
Feedback Systems

Applied Science
Seunghun Kim, GrahamWakefield Juhan Nam

Audio feedback is defined as a positive feedback of acoustic signals where an audio
input and output form a loop, and may be utilized artistically. This article presents
new context-based controls over audio feedback, leading to the generation of desired
sonic behaviors by enriching the influence of existing acoustic information such
as room response and ambient noise. This ecological approach to audio feedback
emphasizes mutual sonic interaction between signal processing and the acoustic
environment. Mappings from analyses of the received signal to signal-processing
parameters are designed to emphasize this specificity as an aesthetic goal. Our feed-
back system presents four types of mappings: approximate analyses of room rever-
beration to tempo-scale characteristics, ambient noise to amplitude and two different
approximations of resonances to timbre. These mappings are validated computation-
ally and evaluated experimentally in different acoustic conditions.

63

Project
학술활동

국제저널

Data-guided Model Predictive Control Based on
Smoothed Contact Dynamics

Computer Graphics Forum
Daseong Han, Haegwang Eom, Junyong Noh, and Joseph S. Shin
(formerly Sung Yong Shin)

In this paper, we propose an efficient data-guided method based on Model Predictive
Control (MPC) to synthesize a full-body motion. Guided by a reference motion, our
method repeatedly plans the full-body motion to produce an optimal control policy
for predictive control while sliding the fixed-span window along the time axis.
Based on this policy, the method computes the joint torques of a character at every
time step. Together with contact forces and external perturbations if there are any,
the joint torques are used to update the state of the character. Without including
the contact forces in the control vector, our formulation of the trajectory optimi-
zation problem enables automatic adjustment of contact timings and positions for
balancing in response to environmental changes and external perturbations. For ef-
ficiency, we adopt derivative-based trajectory optimization on top of state-of-the-art
smoothed contact dynamics. Use of derivatives enables our method to run much
faster than the existing sampling-based methods. In order to further accelerate the
performance of MPC, we propose efficient numerical differentiation of the system
dynamics of a full-body character based on two schemes: data reuse and data inter-
polation. The former scheme exploits data dependency to reuse physical quantities
of the system dynamics at near-by time points. The latter scheme allows the use of
derivatives at sparse sample points to interpolate those at other time points in the
window. We further accelerate evaluation of the system dynamics by exploiting the
sparsity of physical quantities such as Jacobian matrix resulting from the tree-like
structure of the articulated body. Through experiments, we show that the proposed
method efficiently can synthesize realistic motions such as locomotion, dancing,
gymnastic motions, and martial arts at interactive rates using moderate computing
resources.

64

Project
학술활동

국제저널

Influence of control logic on variation of indoor thermal
environment for residential buildings

Indoor and Built Environment
Jin Woo Moon, Ji–Hyun Lee and Sooyoung Kim

This study proposes an advanced thermal control method that employs artificial neural network
(ANN) models for predictive and adaptive thermal control. Two predictive and adaptive con-
trol logic approaches were proposed to simultaneously control indoor temperature and humid-
ity as well as predicted mean vote (PMV) in a residential building. Their thermal performance
was analysed and compared with that of non-ANN-based counterparts to evaluate architectural
variables such as envelope insulation and building orientation. A numerical computer simula-
tion method was used for the tests after demonstration of its validity based on comparison
with results of field measurement. Analysis results revealed that the proposed predictive and
adaptive control methods conditioned the indoor temperature, humidity and PMV effectively.
The periods during which each thermal factor was in a comfortable range increased, and over-
shoots and undershoots out of the targeted comfortable ranges were reduced when using the
ANN model. The results demonstrate the functionality of the proposed method for variation in
architectural variables and that the ANN model has the potential to be successfully applied to
building thermal controls.

65

Project
학술활동

국제저널

Investigating cultural uniqueness in theme parks through
finding relationships between visual integration of visitor
traffics and capacity of service facilities

International Journal of Architectural Computing
Kyung Hoon Hyun, Aram Min, Sun-Joong Kim and Ji-Hyun Lee

The goal of the article is to find the relationships between theme park visitor traffics
and service facility location along with their capacities. To do that, we analyzed four
Disneylands situated in Paris, Tokyo, and the United States (Florida and Califor-
nia). By analyzing the visual integration of visitor traffics at each Disneyland and
calculating the capacities of service facilities such as the attractions, shops, and res-
taurants, we ran through a linear and a geographically weighted regression analysis.
Our results indicate that there is a unique relationship between the service facility
placements and the amount of predicted traffic flows for each Disneyland.

66

Project
학술활동

국제저널

Metaphoric Hand Gestures for Orientation-Aware VR
Object Manipulation With an Egocentric Viewpoint

Transactions on Human-Machine Systems

Youngkyoon Jang, Ikbeom Jeon, Tae-Kyun Kim, and Woontack Woo

We present a novel natural user interface framework, called Meta-Gesture, for se-
lecting and manipulating rotatable virtual reality (VR) objects in egocentric view-
point. Meta-Gesture uses the gestures of holding and manipulating the tools of daily
use. Specifically, the holding gesture is used to summon a virtual object into the
palm, and the manipulating gesture to trigger the function of the summoned virtual
tool. Our contributions are broadly threefold: 1) Meta-Gesture is the first to perform
bare hand-gesture-based orientation-aware selection and manipulation of very small
(nail-sized) VR objects, which has become possible by combining a stable 3-D palm
pose estimator (publicly available) with the proposed static–dynamic (SD) gesture
estimator; 2) the proposed novel SD random forest, as an SD gesture estimator can
classify a 3-D static gesture and its action status hierarchically, in a single classi-
fier; and 3) our novel voxel coding scheme, called layered shape pattern, which is
configured by calculating the fill rate of point clouds (raw source of data) in each
voxel on the top of the palm pose estimation, allows for dispensing with the need
for preceding hand skeletal tracking or joint classification while defining a gesture.
Experimental results show that the proposed method can deliver promising per-
formance, even under frequent occlusions, during orientation-aware selection and
manipulation of objects in VR space by wearing head-mounted display with an at-
tached egocentric-depth camera (see the supplementary video available at: https://
sites.google.com/site/fingergesture/fui/).

67

Project
학술활동

국제저널

Online Real-time Locomotive Motion
Transformation Based on Biomechanical Observations

Computer Animation and Virtual Worlds

Daseong Han, Seokpyo Hong, Junyong Noh, Xiaogang Jin Joseph S. Shin
(formerly Sung Yong Shin)

In the paper, we present an online real-time method for automatically transforming
a basic locomotive motion to a desired motion of the same type, based on biome-
chanical results. Given an online request for a motion of a certain type with desired
moving speed and turning angle, our method first extracts a basic motion of the same
type from a motion graph, and then transforms it to achieve the desired moving
speed and turning angle by exploiting the following biomechanical observations:
The motion of the center of mass (COM) is controlled only by foot contacts with
the ground. The changes of moving speed and direction are accompanied by their
respective forward and lateral leaning of the whole body. Turning is accompanied
by a series of orientation changes of the upper body segments in a top to bottom
sequence from head to pelvis while decreasing moving speed according to the
amount of turning. Exploiting these observations, we propose a simple but effective
method to add physical and behavioral naturalness to the resulting locomotive
motions without preprocessing. Through experiments, we show that our method
enables a character to respond agilely to online user commands while efficiently
generating walking, jogging, and running motions with a compact motion library.
Our method can also deal with certain dynamical motions such as forward roll.

68

Project
학술활동

국제저널

Retargeting Human-Object Interaction
to Virtual Avatars

IEEE Transactions on Visualization and Computer Graphics
Yeonjoon Kim, Hangil Park, Seungbae Bang, and Sung-Hee Lee

Rapid advances of AR technologies are realizing a vision in which people in different
locations can interact seamlessly with each other in a shared space. Recently, there has
been considerable research on virtual avatar-based 3D telepresence in which a user wear-
ing a head mounted display communicates with the image of a remote person overlaid on
the local space of the user [20, 21]. In this case, the physical space that a user belongs to
becomes the telepresence space: The local space is augmented with the remote person’s
avatar, which moves in the local space by following the remote person’s motions made in
the remote place. In addition, as envisioned in [13], we can even make a mirrored telep-
resence space in the remote space by sending the local user’s avatar to the remote place.

If the shapes of the local and remote environments are identical, we can generate the
avatar’s motion simply by copying that of the remote user. Otherwise, the avatar’s mo-
tion needs to be created by adapting the remote user’s motion to the local environment.
In particular, a challenging problem arises when the remote user engages in a physical
interaction (e.g., contact) with an object. Let us assume that there exists a local object
corresponding to the remote object. If the two objects have different shapes, how can we
transfer the remote user’s motion to the avatar? For example, let us assume that the two
spaces are equipped with chairs with different shapes. When the remote user sits on a
chair, the avatar in the local space should be able to perform a sitting motion appropriate
for the shape of the local chair.

In this paper, we confront the problem of retargeting a motion interacting with an object
to a different object that has a similar but different shape. Such motion retargeting should
preserve two features: one is the context of the motion with respect to the object; the other
is the original pose of the human. The difficulty of motion retargeting increases with the
dissimilarity of the function and shape of the two objects. For instance, retargeting a sit-
ting motion on an office chair to a sofa is easier than it would be to the saddle of a bicycle.
We introduce a novel motion retargeting technique for two objects that have similar func-
tions and shapes. More specifically, we assume that there is a correspondence between
the surfaces of the objects such that for every point on one surface there exists a cor-
responding point on the other, and that an arbitrary continuous curve on one surface can
be mapped to a continuous curve on the other. In other words, it is assumed that a continu-
ous and bijective map f : S1 →S2 exists between the two surfaces Si. This assumption
allows for a continuous movement of contact points between a user and an object to be
transferred to a continuous contact motion on the other object. Then, how can a motion
close to yet not in contact with an object be transferred to the nearby space around the
other object? This problem has been more or less overlooked except for a few studies,
e.g., [2, 10, 24].

69

Project
학술활동

국제저널

Rich360: Optimized Spherical Representation from
Structured Panoramic Camera Arrays

ACM Transaction on Graphics
Jungjin Lee, Bumki Kim, Kyehyun Kim, Younghui Kim, Junyong Noh

This paper presents Rich360, a novel system for creating and viewing a 360 pano-
ramic video obtained from multiple cameras placed on a structured rig. Rich360
provides an as-rich-as-possible 360 viewing experience by effectively resolving two
issues that occur in the existing pipeline. First, a deformable spherical projection
surface is utilized to minimize the parallax from multiple cameras. The surface is
deformed spatio-temporally according to the depth constraints estimated from the
overlapping video regions. This enables fast and efficient parallax-free stitching in-
dependent of the number of views. Next, a non-uniform spherical ray sampling is
performed. The density of the sampling varies depending on the importance of the
image region. Finally, for interactive viewing, the nonuniformly sampled video is
mapped onto a uniform viewing sphere using a UV map. This approach can preserve
the richness of the input videos when the resolution of the final 360 panoramic video
is smaller than the overall resolution of the input videos, which is the case for most
360 panoramic videos. We show various results from Rich360 to demonstrate the
richness of the output video and the advancement in the stitching results.

70

Project
학술활동

국제저널

SketchiMo: Sketch-based Motion Editing for Articu-
lated Characters

ACM Transaction on Graphics
Byungkuk Choi, Roger Blanco i Ribera, J. P. Lewis, Yeongho Seol.
Seokpyo Hong, Haegwang Eom, Sunjin Jung, Junyong Noh

We present SketchiMo, a novel approach for the expressive editing of articulated
character motion. SketchiMo solves for the motion given a set of projective con-
straints that relate the sketch inputs to the unknown 3D poses. We introduce the
concept of sketch space, a contextual geometric representation of sketch targets —
motion properties that are editable via sketch input— that enhances, right on the
viewport, different aspects of the motion. The combination of the proposed sketch
targets and space allows for seamless editing of a wide range of properties, from
simple joint trajectories to local parent-child spatiotemporal relationships and more
abstract properties such as coordinated motions. This is made possible by interpret-
ing the user’s input through a new sketch-based optimization engine in a uniform
way. In addition, our view-dependent sketch space also serves the purpose of disam-
biguating the user inputs by visualizing their range of effect and transparently defin-
ing the necessary constraints to set the temporal boundaries for the optimization.

71

Project
학술활동

국제저널

Sparse feature learning for instrument identifica-
tion: Effects of sampling and pooling methods

Journal of the Acoustical Society of America
Yoonchang Han, Subin Lee, Juhan Nam, and Kyogu Lee

Feature learning for music applications has recently received considerable attention
from many researchers. This paper reports on the sparse feature learning algorithm
for musical instrument identification, and in particular, focuses on the effects of
the frame sampling techniques for dictionary learning and the pooling methods for
feature aggregation. To this end, two frame sampling techniques are examined that
are fixed and proportional random sampling. Furthermore, the effect of using onset
frame was analyzed for both of proposed sampling methods. Regarding summari-
zation of the feature activation, a standard deviation pooling method is used and
compared with the commonly used max- and average-pooling techniques. Using
more than 47 000 recordings of 24 instruments from various performers, playing
styles, and dynamics, a number of tuning parameters are experimented including the
analysis frame size, the dictionary size, and the type of frequency scaling as well as
the different sampling and pooling methods. The results show that the combination
of proportional sampling and standard deviation pooling achieve the best overall
performance of 95.62% while the optimal parameter set varies among the instru-
ment classes.

72

Project
학술활동

국제저널

TunnelSlice: Freehand Subspace Acquisition Using
an Egocentric Tunnel for Wearable Augmented Reality

Transactions on Human-Machine Systems
Hyeongmook Lee, Seung-Tak Noh, and Woontack Woo

In this paper, we propose TunnelSlice, which enables natural acquisition of sub-
space in an augmented scene from an egocentric view, even for scenarios involving
ambiguous center objects or object occlusion. In wearable augmented reality (AR),
approaching a three-dimensional (3-D) region including the objects of interest has
become more important than approaching distant objects one by one. However, ex-
isting ray-based volumetric selection through a head worn display accompanies dif-
ficulties in defining a desired 3-D region due to obstacles by occlusion and depth
perception. The proposed TunnelSlice effectively determines a cuboid transform,
excluding unnecessary areas of a user-defined tunnel via two-handed pinch-based
procedural slicing from an egocentric view. Through six scenarios involving cen-
tral object status and different occlusion levels, we conducted a user study of Tun-
nelSlice. Compared with two existing approaches, TunnelSlice was preferred by
the subjects and showed greater stability for all scenarios, and outperformed the
other approaches in a scenario involving strong occlusion without a central object.
TunnelSlice is thus expected to serve as a key technology for spatial protocol and
interaction using a subspace in wearable AR.

73

Project
학술활동

국제저널

Virtual Reality, Augmented Reality and Augmented Human

Journal of Virtual Reality Society of Japan
Jeongmin Yu, Woontack Woo

In this paper, we briefly overview the past and recent research activities in the areas
of Virtual Reality (VR) and Augmented Reality (AR) in Korea. We first briefly sum-
marize major VR/ AR research activities for last ten years and then introduce the re-
cent research activities laying emphasis on research conducted in the Ubiquitous VR
(UVR) Laboratory at KAIST. Finally, we share our vision on ‘UVR and Augmented
Human (AH)’ and possible future direction [2]. The VR/AR research in Korea and
its trends were not much different with the global trend. In Korea, research in AR/
VR field was initiated in the late 1980s and early 1990s by pioneers including
Professors Kwangyun Won (KAIST), Prof. Hyun Seung Yang (KAIST), Profes-
sor Myoung-Hee Kim (Ewha Womans University), Dr. Heedong Ko (KIST), etc.
Their research results have been shared in HCI Korea conference since 1991. Since
mid-90s various VR/AR R&Ds had been actively conducted in many Universi-
ties (POSTECH, Kyungpook National University, etc.), Research Institutes (KIST,
ETRI, etc.) and Industries (Samsung, Hansaem, etc.). Unfortunately, however, the
R&D activities in Industries had fallen into ‘trough of disillusionment’ in the early
2000s and lasted until the mid-2010s since the R&D results could not meet the ex-
pectations of the market.

74

Project
학술활동

국내학회

Gamification Can Affect the Sustainability of
The Wearable Device for Physical Activity?
: User Case Study

HCI 2016
Seukwoo Lee, Jaemyung Lee, Sangeun Kum, Young Yim Doh

Recently, wearable device market is popular worldwide. Among several kinds of it, one area is
pedometer-based fitness service and products. Previous studies have mainly focused on explor-
ing the technological development of it. However, while gradually increasing the interests in
user’s usage patterns, game and gamification studies have shown positive results to promote
user engagement of physical activity. The purpose of this study is to see the usability and
sustainability on the wearable device for physical activity. We want to observe effectiveness
specially in gamification conditions through user case study. We performed user evaluations
of Fitbit, one of the pedometer-based wearable devices and FitRPG, a gamified service compat-
ible with Fitbit. Interview questions consist of usability, usefulness and affect based on Human
Computer Interaction (HCI) elements. We conducted user research for 4 weeks (2 weeks for
Fitbit condition/2 weeks for FitRPG condition) to ascertain sustainability. After collecting data
and interview answers, we were able to suggest three issues of the gamified service that may
affect users.

75

Project
학술활동

국내학회

머신러닝 기반 소셜 데이터 모니터링 플랫폼 연구
- 일본 원전 사고 이후의 원자력 관련 담화를 중심으로 -

한국정보과학회
하유이, 최지정, 차미영, 이지연, 김병직, 이동명
2011년 일본 후쿠시마 원자력 발전소 사고 이후 국내 소셜네트워크 서비스에서는
원자력에 관한 다양한 담론이 진행되고 있다. 이 연구에서는 사고 전후 약 4년간
원자력을 언급하는 15만건의 트위터 데이터를 수집하고 핵심 키워드와 의미망을
추출 및 분석하였다. 정보 전파를 관찰하기 위해 원자력 관련 정부의 공식기관,
뉴스 미디어 및 일반 대중의 담론을 TF-IDF 기법과 나이브 베이지안 기반으로 분
석 및 분류하였다. 그 결과 ‘오염’에 관한 토픽을 중심으로 담론이 발전하였으
며, 세가지 언급 주체 간의 중심 토픽의 차이와 상관관계를 확인할 수 있었다. 이
러한 모니터링 플랫폼은 향후 재난 위기 상황 발생 시 효과적인 대중 커뮤니케이
션 체계 구축에 활용될 수 있다.

76

Project
학술활동

국내학회

Measuring creativity in music via a network analy-
sis of codeword transitions

NetSci 2016
Doheum Park and Juyong Park

Creativity has long been studied by multidisciplinary scholars but its secret
still remains hidden.The recent advent of datasets for the arts allows us to
identify universal patterns in creative works and understand the signicant
features of creativity through computational models. Our work applies the
network framework to the classical music MIDI dataset that spans several
centuries to explore a computational model of creativity and understand the
evolution of music. Our model of creativity in music utilizes two concepts,
P-creativity and H-creativity, suggested by Boden [1]; P-creativity computes
how novel the piece is with respect to the previous works by the same com-
poser, while H-creativity evaluates how novel it is with respect to all the
previous works by all the composers. To calculate the creativity of each
piece, we build a Markov network of \codeword transitions,” the changes of
simultaneous pitches, as shown in Fig.

77

Project
학술활동

국내학회

360도 비디오를 위한 몰입형 디스플레이

Korea Computer Graphics Society 2016
서형국, 김재동, 노준용
최근 대표적인 가상현실(Virtual Reality) 콘텐츠중 하나인 360도 비디오는 인터
넷에서 손쉽게 접할 수 있고 심지어 모바일 기기를 이용해 일반 사용자들도 손쉽게
제작 할 수 있게 되었다. 이러한 360도 비디오를 재생할수 있는 시스템은 현재 일
반 모니터, 모바일 기기, 그리고 Head Mounted Display(HMD)등이 있다. 그러나 전
통적인 모니터 디스플레이나 모바일 기기는 충분한 몰입감을 제공하기에 적합하지
않으며, HMD의 경우 훌륭한 몰입감을 제공할 수는 있지만, 기기를 머리에 장착해
야 하는 불편함을 수반한다. 본 연구에서는 360도 비디오를 재생하기 위한 새로운
수단으로 다면 상영 시스템 기반의 몰입형 디스플레이를 제안하고자 한다.

78

Project
학술활동

국내학회

The Judge-Contestant Network in Competitions:
Biases and Clusters from the 17thInternational
Chopin Piano Competition

NetSci 2016
Gyuhyeon Jeon and Juyong Park

A common form of competition in culture is one in which a jury grades the
competitors’ perfor-mances. Unlike in refereed one-on-one sports, it is
dicult to gauge the role of subjectivity in the grading which may some-
times give rise to a controversy surrounding an outcome. For instance, in
the 17th International Chopin Piano Competition took place in 2015, French
judge Philippe Entremont (abbreviated PE) was accused of baselessly giving a
low score to Sung-jin Cho (SC), who went on to win the general competition.
We apply the hierarchical clustering method to the jury-contestant scoring
network to validate the public’s perception of his unfairness; Our basic
hypothesis is that if he had been unfair to a specic contestant, a hypotheti-
cal absence of the contestant would result in signicantly dierent clustering
for the jury. In Figure 1 (a) and (b) we show the dendrograms of the judges
based on cosine similarity.

79

Project
학술활동

국내학회

The Multi-Scale Network Landscape of
Collaboration

NetSCI 2016
Arram Bae, Doheum Park, Yong-Yeol Ahn, Juyong Park

Propelled by the increasing availability of large-scale high-quality data,
advanced data modeling and analysis techniques are enabling many novel and
signicant scientic understanding of a wide range of complex social, natural,
and technological systems. These developments also provide opportunities for
studying cultural systems and phenomena which can be said to refer to all-
products of human creativity and way of life. An important characteristic of
a cultural product is that it does not exist in isolation from others, but
forms an intricate web of connections on many levels. In the creation and
dissemination of cultural products and artworks in particular, collaboration
and communication of ideas play an essential role, which can be captured in
the heterogeneous network of the creators and practitioners of art. In this
paper we propose novel methods to analyze and uncover meaningful patterns
from such a network using the network of western classical musicians con-
structed from a large-scale comprehensive Compact Disc recordings data. We
characterize the complex patterns in the network landscape of collaboration
between musicians across multiple scales ranging from the macroscopic to the
mesoscopic and microscopic. On the macroscopic scale we see massive inequal-
ity in importance of musicians. On the mesoscopic scale we witness emergence
of schools re ecting styles and origins. Finally, on the microscopic scale,
we uncover the signicance of individuals that can be buried on other scales.
Our work establishes two hallmark properties of culture, the diversity of
cultural styles and the individuality of the artists, as representing the
cultural landscape on dierent scales.

80

Project
학술활동

국내학회

XML을 이용한 디지털 건축문화유산 콘텐츠 저작 도구

2016년 한국컴퓨터정보학회 동계학술대회
이종욱, 이지형, 이지현
전문적인 건축 정보가 제공되는 건축문화유산 콘텐츠를 위해서는 콘텐츠에서의 디
지털 문화유산은 실재 문화유산의 구조 및 구성을 반영하여야 한다. 즉, 건축문화
유산의 구성 요소인 부재를 기반으로 제작된 디지털 문화유산이 콘텐츠에서 제공
되어야 한다. 본 연구에서는 XML기반의 건축문화유산 정보구조를 정의하고 이를
기반으로 한 콘텐츠 저작도구 및 체험 콘텐츠를 소개한다.

81

Project
학술활동

국내학회

공공 인터랙티브 예술 작품 “deBallution” 제작을 위한 기초
프로토타입 디자인 연구

HCI KOREA 2016
오제호, 김소영, 박성일, 남윤태, 시정곤

이번 연구는 미디어 파사드에 대한 공공 인터랙티브 예술 작품 “deBallution”
구현을 위한 프로토타입 제작 기초 디자인을 하는 것이다. “deBallution”의 기
본 콘셉트는 미디어 파사드에서 도시의 풍경을 재현하는 영상이 재생되고 관객들
은 미디어 파사드에 대고 던지는 행동을 계속 수행한다. 이 행동 작용에 의해서
키넥트와 오픈프레임웍스를 사용해서 영상 콘텐츠에 만화경의 이미지가 재생되기
시작하고 점점 주변부로 확장된다. 관객들의 계속된 행동에 의해 미디어 파사드
속 만화경 이미지가 다 채워지면 기존의 도시의 풍경 영상 콘텐츠가 전복되고 관
객들은 정서적 변화를 경험하게 된다. 이번 연구에서는 “deBallution”의 구현
을 위한 프로토타입의 기초 디자인으로 사용자 시나리오 디자인, 그래픽 디자인,
테크니컬 디자인을 제안했다.

82

Project
학술활동

국내학회

게임 내 가상공간 분류 기준 확립 및 대표 유형 산출

한국게임학회 2016 추계학술대회
김익환, 이인정, 이지현

컴퓨터 기술이 발전함에 따라 게임 속 공간들이 다양한 모습들로 구현되고 있다.
공간 유형들의 종류가 다양하고 광범위한 탓에 공학적인 설계 방법론을 구축하고
자 하여도 그 영역을 설정하기에는 여태 어려움이 있었다. 이에 본 연구는 이러
한 게임 내의 공간들을 분류하는 다섯 가지 기준들을 제시하였으며 해당 분류 기
준과 방법에 대한 유효성 평가를 진행하였다. 해당 다섯 가지 분류 기준들은 각각
공간의 서술특성, 공간과 동작차원, 서술형태, 협업 정도, 상호교환 정도이다. 무
작위 산출한 각각 50개씩의 PC와 콘솔 게임들을 대상으로 해당 기준으로 분류 작
업을 진행한 결과, 예외 없이 전량 분석이 가능하였음에 다섯 가지 기준의 유효성
을 검증할 수 있었다. 확보된 기준들을 바탕으로 1996년부터 2016년까지 9월까지
연도별 가장 주목 받은 게임 총 385개를 다섯 가지 기준들을 토대로 분류하여 그
변화 추이를 살펴보았으며, 세가지 대표 유형들의 산출이 가능하였다. 해당 분류
방법론은 향후 게임 속 가상경관 설계 방법론을 구축함에 있어 건축 및 조경의 영
역에서 활용되고 있는 각종 설계 방법론의 적용이 가능하게끔 유도하는 역할을 수
행할 것이며, 근본적으로 게임 공간을 설계하고 구현하는 인원들에게 보다 체계적
이고 효율적인 접근을 제공할 것이다.

83

Project
학술활동

국내학회

딥러닝을 활용한 조기 루머 탐지 기법

동계학술대회
권세정

온라인 미디어를 통한 루머의 전파는 빠른 전파력으로 인해 조기 탐지의 필요성
이 지속적으로 야기되는 이슈이다. 이 연구에서는 트위터 API에서 제공하는 초
기 트윗만을 활용했을 때 루머의 탐지가 가능한지를 딥러닝 기법 중 하나인 RNN
(Recurrent Neural network)를 통해 확인하였다. 498개의 루머와 494개의 비루머
이벤트들은 초기 트윗과 연관된 사용자 특성 (팔로워 수, 친구수) 및 반응 특성
(favorite 수, 리트윗수)에 기반한 다변량 시계열 자료로서 표현된다. 주요 변수
인 초기 트윗의 개수는 20개에서 시작해 100개까지로 통제하여 실험을 진행하였
으며, 30개의 초기 트윗 (평균 관찰 기간 4 시간)만으로 약 0.77의 F1 점수의 루
머 구분력을 보였다. 이것은 기존의 루머 구분 연구보다 월등히 짧은 시간 및 효
율적인 전처리만으로도 경쟁력 있게 루머를 구분함을 의미한다. 이 연구의 결과
는 향후 보다 빠르고, 효율적인 루머 탐지 시스템 구축을 위한 주요 기반 연구로
서 활용 가능하다.

84

Project
학술활동

국내학회

비디오 게임 속 가상경관 분류 기준 제시 및 유효성 증명
- PC 및 Console 게임을 기준으로 -

2016한국조경학회 추계학술대회
김익환, 이기쁨, 이지현

아동, 청소년들의 온라인 게임 이용이 보편화되면서, 자녀들의 건강한 온라인 게
임 이용을 위한 부모의 역할이 강조되고 있다. 본 연구에서는 부모들의 온라인 게
임 인식 유형의 특성을 분석하고, 이를 근거로 부모들이 균형 잡힌 온라인 게임
인식을 가질 수 있도록 하기 위한 유형별 개입 전략을 제안하였다. 이를 위해 초
등학생부터 대학생 자녀를 둔 부모들을 대상으로 온라인 게임에 대한 인식 및 관
련 변인을 설문조사를 통해 조사하였고, 총 345명의 자료가 분석되었다. 분석 결
과, 온라인 게임 인식 척도를 통해 부모의 온라인 게임 인식은 4유형으로 구분되
었고, 각 유형은 자녀의 학업에 미치는 영향, 자녀의 게임 과몰입에 대한 지각,
온라인 게임 친숙도 및 인터넷 리터러시 등에서 서로 다른 특성을 보였다. 본 연
구에서는 각 유형별 특성에 근거하여 균형 잡힌 온라인 게임 인식 형성을 위한 개
선 지점과 구체적인 전략의 예를 제안하였다. 본 연구는 획일화된 정책이 아닌 유
형별 특성에 근거한 맞춤형 전략을 제시하였다는데 의의가 있다.

85

Project
학술활동

국내학회

사용자 컨텐츠 기반 데이터를 이용한 소셜 미디어 타겟팅 광고
연구: 패션 브랜드의 인스타그램 네이티브 광고
중심으로

2016 기술경영경제학회 하계학술대회
김희철, 차미영, 김원준

패션 제품에 대한 온라인 소비 및 수요가 급증하는 시대이다. 온라인 매출 성장
률이 급성장하고 있는 패션업계의 변화에 발맞춰 일반적인 다수의 사용자에게 저
비용으로 광고를 노출하는 방법 외에 소셜 미디어 데이터 분석을 이용해 잠재 구
매자에게 전략적인 온라인 광고 설계에 대한 필요성도 함께 증가했다.
기존 이론에서는 온라인 광고가 일반 사용자층이 아닌 온라인 구매자에게 노출됐
을 경우 그 효과가 더 높았으며, 이러한 온라인 구매자들에게 노출된 광고는 그들
의 스타일 인스피레이션이 되어 언젠가 구매로 이어질 가능성이 훨씬 높음을 제시
하고 있다. 따라서 본 논문에서는 광고의 단가를 낮추는 전략이 아닌 브랜드를 직
접 언급하는 잠재 구매자에게 집중 타겟팅하는 전략을 중심으로, 실제 소셜 미디어
데이터를 활용하여 타겟 브랜드와 함께 언급된 브랜드들을 선호하는 온라인 사용
자들에게 집중 타겟팅한 광고가 효율적이라는 가설을 검증하고자 했다.
이를 위해 일반적인 온라인 사용자가 아닌 잠재 온라인 구매 고객에게만 집중적
으로 광고를 노출시키는 타겟팅 광고를 시도해봤다. 구체적으로는 타겟 브랜드와
함께 언급된 다른 주요 브랜드를 선정하여 해당 브랜드에 관심을 표명한 온라인 사
용자 대상으로 광고를 타겟팅했다. 이는 타겟 브랜드에 개인화된 광고 전략을 찾는
시도라고 할 수 있겠다.
본 연구의 결과는 다음과 같다. 타겟 브랜드와 공동으로 언급이 된 브랜드들이
집중 타겟팅에 사용될 경우 광고에 대한 더욱 적극적인 반응을 확인할 수 있었다.
뿐만 아니라 온라인 광고 자체의 실질적인 효능은 단지 웹사이트에서 보이는 광고
집행의 지표를 넘어서는 브랜드 이미지 상승 및 구매 촉진 등 복합적인 지표를 살
펴야 한다는 시사점도 제시할 수 있겠다.

86

Project
학술활동

국내학회

일상 생활에서 신체 활동을 증진하기 위해
게임화와 상징화 전략을 활용한 모바일 게임 디자인 연구

HCI 2016
이석우, 이주영, 서기슬, 도영임

이 논문에서는 일상 생활에서 신체 활동을 증진하기 위해 게임화와 상징화 전략을
활용한 모바일 헬스게임 디자인을 제시하고자 한다. 이를 위해 디자인 방법론을
적용한 프로토타입을 구상 및 설계 하였다. 일상 생활에서 걷기 활동을 증진하기
위해, 한국전통 놀이인 땅 따먹기의 상징 구조를 응용한 모바일 헬스 게임을 안
드로이드 스마트폰을 사용하여 개발하였다. 6 명의 실험 참가자를 대상으로 파일
럿 테스트를 수행하고 포커스 그룹 인터뷰(Focus group interview)를 통해 개발한
프로토타입의 만족도와 디자인 시사점을 확인하였다. 연구 결과 개발된 프로토
타입을 통해 위치 좌표 인식 방식 알고리즘과 사용자들의 멘탈 모델 차이, 사용
자의 행동이 적절한 피드백을 받지 못할 때의 동기 부여 저하, 상징화와 시각적
피드백의 중요성, 그리고 경쟁 및 사회적 소통관계의 게임화에 대한 제언 등 4
가지 디자인 시사점을 도출하였다. 이를 바탕으로 후속 연구의 필요성을 제시하
였다.

87

Project
학술활동

국내학회

행위 맥락에 따른 장소의 감성 분석:워드 임베딩을 이용하여

제43회 정기총회 및 동계학술발표회 (2016, 한국정보과학회)
민세미, 박주용

Case Study of Fashion Brand’s
Instagram Hashtag Co-occurrence Analysis

NetSci 2016
Heechul Kim, Meeyoung Cha

Fit Friends: The Importance of a Supportive Social Network
for Persistent Fitness Sharing

NetSci 2016
KunwooPark, Ingmar Weber, MeeyoungCha, and ChulLee

Identifying popular songs’ success through the lens of their
chart dynamics

NetSci 2016
Seungkyu Shin and Juyong Park

Korea Public Perception on Fukushima Nuclear Accident

NetSci 2016
Seung-Hoi Kim, Yu-i Ha, Meeyoung Cha, Jiyon Lee, Byung-Jik Kim, Dong-Myung Lee

Nowcasting Commodity Prices using Social Media

NetSci 2016
JaewooKim, MeeyoungCha, Jonggun Lee

Prominent Features of Rumor Propagation in
Online Social Media

NetSci 2016
Sejeong Kwon, Meeyoung Cha, Kyomin Jung, Wei Chen, Yajun Wang

88

Project
학술활동

국제학회

“Vernacular Sound”: System for Soundscaping of Everyday
Objects

International Symposium on Electronic Art
JooYoung Oh, JuYoung Lee, YouJin Lee, KwangYun Won

Creating sound in diverse spaces to generate timbre without limitations and enhance
experience in soundscape has been an essential issue for sound artists. We used con-
cept of ‘Vernacular’ as an appropriation for Soundscape of everyday object, allow-
ing listener to compete the current sound consuming environment which eliminates
noise during the process of sound recording. We reviewed previous works on sound-
scape, noise art and object used in sound art to enhance the sound space experience.
We designed a system for soundscaping, Vernacular Sound, consists of two main
parts: (1) audio effect which computes the sound space of the object and (2) audio
visualization which enables interaction between the listener and the sound space via
Kinect. With ‘Vernacular Sound’ system, our purpose is to provide new experience
of listening to sound using everyday objects.

A Unified Framework for Remote Collaboration Using
Interactive AR Authoring and Hands Tracking

HCII 2016
Jeongmin Yu, Jin-u Jeon, Gabyong Park, Hyung-il Kim, Woontack Woo

In this paper, we present a unified framework for remote collaboration using interac-
tive augmented reality (AR) authoring and hand tracking methods. The proposed
framework enables a local user to organize AR digital contents for making a shared
working environment and collaborate multiple users in the distance. To develop the
framework, we combine two core technologies: (i) interactive AR authoring method
utilizing a smart input device for making a shared working space, (ii) hand-augment-
ed object interaction method by tracking two hands in egocentric camera view. We
implement a prototype of the proposed remote collaboration framework for testing
its feasibility in an indoor environment. To the end, we expect that our framework
enables collaboration as feeling a sense of co-presence with remote users in a user’s
friendly AR working space.

89

Project
학술활동

국제학회

All-in-One Mobile Outdoor Augmented Reality Framework
for Cutural Heritage Site

Signal-Image Technology & Internet-Based System
Noh-young Park, Eunseok Kim, Jongwon Lee, Woontack Woo

In this paper, we propose an all-in-one mobile outdoor augmented reality (AR)
framework for a cultural heritage site. The framework was designed to incorporate
computer vision-based augmented reality technology and ontology-based data-
authoring technology. Through this framework, we clearly explain how to create
3D visual data for camera pose estimation and how to connect AR content with a
cultural heritage site. In addition, we suggest a multithreading camera tracking and
offer an estimation model for mobile AR application. Finally, we have confirmed the
efficiency and reliability of our framework. Through this vision-based AR frame-
work, seamless AR application for a cultural heritage site can be made.

An Eulerian approach for constructing a map between sur-
faces with different topologies

Pacific Graphics 2016 Conference
Hangil Park, Youngjin Cho, Seungbae Bang, Sung-Hee Lee

3D objects of the same kind often have different topologies, and finding correspond-
ence between them is important for operations such as morphing, attribute transfer,
and shape matching. This paper presents a novel method to find the surface corre-
spondence between topologically different surfaces. The method is characterized by
deforming the source polygonal mesh to match the target mesh by using the inter-
mediate implicit surfaces, and by performing a topological surgery at the appropriate
locations on the mesh. In particular, we propose a mathematically well-defined way
to detect the topology change of surface by finding the non-degenerate saddle points
of the velocity fields that tracks implicit surfaces. We show the effectiveness and
possible applications of the proposed method through several experiments.

90

Project
학술활동

국제학회

An Intuitive Heritage Education System for Learning
Architectural Structures and Styles

the 34th eCAADe Conference
Jongwook Lee, Aram Min, Jihyun Lee

Although national and international attention toward the cultural heritage is on a
rise, there is a lack of public attention toward cultural heritage sites. One of the main
problems is the lack of interaction and the visualization of cultural heritage on the
sites. Especially Korean historical architectures are structurally complex and are not
easily understood by common people. To improve the public awareness of architec-
tural heritages, we propose a model-based diagnosis system to educate the visitors
and tourists. The system is designed to guide the users to model an appropriate
architecture in accordance with the era, location, and the usage. For the system, we
built a robust set of cases based on the ontological structure we designed especially
for architectural heritage education. It basically enables users to reconstruct build-
ings intuitively in six steps from bottom to top. A system evaluation was conducted
on the affective, cognitive, operative aspect of the system at a heritage site. The
results show well in terms of cognitive aspect but was evaluated poorly in terms of
the operative aspect.

An Ontology-Based Augmented Reality Application Explor-
ing Contextual Data of Cultural Heritage Sites

Signal-Image Technology & Internet-Based System
Hangil Park, Youngjin Cho, Seungbae Bang, Sung-Hee Lee

91

Project
학술활동

국제학회

Characteristics of Non-linguistic Vocalizations as Auditory
Emoticons

Proceedings of International Conference on
Music Perception and Cognition (ICMPC), 2016
Kyoungsoo Chun, Jinah Kwak, Juhan Nam

Non-verbal communication is important part of interpersonal communication.
Among others, facial expressions and non-linguistic vocalizations (e.g. laughing or
crying) are used as effective means to deliver messages without words or control the
strength of messages in words particularly when emotions are involved. These days,
in computermediated communication (CMC), emoticons have played such a role in
a non-verbal manner. While the text or image-based symbols have been popularly
used, their acoustic counterpart, or non-linguistic vocalizations have been rarely
handled in the CMC. In this paper, we explore the possibility of using non-linguistic
vocalization sounds as an auditory emotion. To this end, we built a dataset of non-
linguistic vocalization sounds and surveyed how the sounds are emotionally well-
paired with emoticons and what characteristics they have in semantic and acoustic
aspects. Specifically, we focused on cuteness, naturalness, childness and gender in
semantic aspect, considering general properties of image and text emoticons. In
acoustic aspect, we analyzed pitch, energy, onset and length. The results show that,
in determining the suitability as an auditory emoticon, different characteristics are
important depending on emotion types. However, those recognized as cute and natu-
ral, and with relatively high pitch or short length, are generally more appropriate as
an auditory emoticon. This semantic and acoustic analysis may be used as a guide
for designing auditory emoticons.

92

Project
학술활동

국제학회

Comparison of Tonality Models in Measuring Chord Se-
quence Similarity

Proceedings of International Conference on
Music Perception and Cognition (ICMPC), 2016
Kyoungsoo Chun, Jinah Kwak, Juhan Nam
The objective of this study is to examine chord sequence similarity measures and
experimentally assess their relationship with human perception of the similarity. We
present five different types of chord sequence similarity measures based on differ-
ent tonality models, including 1) tonal dissonance of intervals, 2) circle of fifths, 3)
harmonic relations, 4) tonal pitch space, and 5) hierarchy of harmonic stability. For
the evaluation, we collected 50 chord sequence pairs from US musical copyright
infringement cases. Also, we surveyed human evaluation to compare it with the
computational chord sequence similarity. The results show that those based on tonal
pitch space and hierarchy of harmonic stability are relatively more correlated with
the human judgement data and a combination of the two similarity measures further
increases the correlation.

93

Project
학술활동

국제학회

CULTURAL DIFFERENCE IN COLOUR USAGES FOR BUILDING
FAÇADES FOCUSING ON THEME PARK BUILDINGS

Conference on Computer-Aided Architecture Design Research
in Asia (CAADRIA)
Eunseok Kim, Junghwa Kim, Woontack Woo

The notion of glocalization has become widely spread in various fields, and accord-
ingly, it is increasingly more important to take account of indigenous culture charac-
teristics in each field. An as-pect of achieving glocalization, globalization with local
consideration, is to consider the difference of colour usage between distinct cultures.
This study suggests an approach to investigate the colour difference between eastern
and western cultures with the case analysis of build-ing façade colours in Disney-
land Paris and Tokyo Disneyland. We an-alysed cultural colour usage characteristics
and derived tendencies for both Paris and Tokyo Disneyland building façade col-
ours. To do this, we use image based k-means clustering algorithm and CIELAB
colour space distances to explore colour characteristics. Our analysis indi-cates an
overall colour usage tendency that Paris uses more green and bluish colours and
Tokyo uses more red and yellowish colours for building façades, based on CIELAB
colour space values. The major motivation of this paper was to reflect the atmos-
phere and the mood of the space that can be easily felt but not readily expressible
into a cultural colour palette. Eventually, by finding the characteristics of perceived
colours, we hope to create a colour recommendation system for different cultures
based on cultural clues.

94

Project
학술활동

국제학회

CulturalHeritageAR: An Ontology-based Outdoor AR
Application for Cultural Heritage Sites

Asian Workshop on User Interface (AUI 2016)
Hayun Kim, Jungwha Kim, Woontack Woo

In this paper we present a new and interesting application to enhance the experience
of visitors to cultural heritage sites. The CulturalHeritageAR is an augmented real-
ity (AR) application that offers contextual information on POIs based on cultural
heritage ontology. By selecting themes and relationships, users of the application
can browse not only information on the POI, but also information related to the POI
including object, event, actor, place, and time.

Data-guided Model Predictive Control Based on
Smoothed Contact Dynamics

the 34th eCAADe Conference
Amartuvshin Narangerel, Ji-Hyun Lee, Rudi Stouffs

A building façade plays an important role of reducing artificial lighting by introduc-
ing natural light into the interior space. A majority of research and current tech-
nology heavily focuses on the optimization of window properties such as the size,
location, and glazing with the consideration of external shading device as well as
the building wall in order to obtain appropriate natural lit space. In the present work,
we propose a 3-dimensional approach that can explore the trade-offs between two
objectives, daylight performance and electricity generation, by means of paramedic
modeling and multi-objective optimization algorithm. The case study was simulat-
ed under the environmental setting of the geographical location of Incheon, Korea
without any urban context. Using the proposed methods, 50 pareto-front optimal
solutions were derived and investigated based on the achieved daylighting and gen-
erated electricity.

95

Project
학술활동

국제학회

Data-guided Model Predictive Control Based on
Smoothed Contact Dynamics

The 12th International Conference on Ubiquitous Robots and Ambient Intelligence (URAI 2015)
Daseong Han, Haegwang Eom, Junyong Noh, Joseph S. Shin
(formerly Sung Yong Shin)

In this paper, we propose an efficient data-guided method based on Model Predictive
Control (MPC) to synthesize a full-body motion. Guided by a reference motion, our
method repeatedly plans the full-body motion to produce an optimal control policy
for predictive control while sliding the fixed-span window along the time axis. Based
on this policy, the method computes the joint torques of a character at every time
step. Together with contact forces and external perturbations if there are any, the
joint torques are used to update the state of the character. Without including the con-
tact forces in the control vector, our formulation of the trajectory optimization prob-
lem enables automatic adjustment of contact timings and positions for balancing in
response to environmental changes and external perturbations. For efficiency, we
adopt derivative-based trajectory optimization on top of state-of-the-art smoothed
contact dynamics. Use of derivatives enables our method to run much faster than
the existing sampling-based methods. In order to further accelerate the performance
of MPC, we propose efficient numerical differentiation of the system dynamics of
a full-body character based on two schemes: data reuse and data interpolation. The
former scheme exploits data dependency to reuse physical quantities of the system
dynamics at near-by time points. The latter scheme allows the use of derivatives
at sparse sample points to interpolate those at other time points in the window. We
further accelerate evaluation of the system dynamics by exploiting the sparsity of
physical quantities such as Jacobian matrix resulting from the tree-like structure of
the articulated body. Through experiments, we show that the proposed method ef-
ficiently can synthesize realistic motions such as locomotion, dancing, gymnastic
motions, and martial arts at interactive rates using moderate computing resources.

96

Project
학술활동

국제학회

“DreamHouse”: NUI-based Photo-realistic AR Authoring
System for Interior Design

Augmented Human International Conference (AHIC) 2016
Jinwoo Park, Sung Sil Kim, Hyerim Park, Woontack Woo

This paper proposes a system which enables users to have enhanced interior deign-
ing and authoring tool through augmented reality (AR). The proposed system, which
we refer to as DreamHouse, focuses on providing natural user interaction and re-
alistic AR experience by enabling following features: 1) allowing users to utilize
bare-hand interaction with an attached egocentric RGB-D camera when trying to
make detail adjustments to virtual objects’ size, location and orientation, 2) render-
ing every virtual object in consideration of an environment and lighting conditions,
creating photo-realistic scenes to help users to have immersive and realistic interior
designing experience. As a result, DreamHouse allows users to freely and easily
interact in physical space with virtual objects using bare hands and gives immersive
and realistic AR authoring experience to users through photo-realistic rendering.

Efficient 3D Hand Tracking in Articulation Subspaces for
the Manipulation of Virtual Objects

Computer Graphics International 2016 (CGI 2016)
Gabyong Park, Antonis Argyros, Woontack Woo

We propose an efficient method for model-based 3D tracking of hand articulations
observed from an egocentric viewpoint that aims at supporting the manipulation
of virtual objects. Previous modelbased approaches optimize non-convex objective
functions defined in the 26 Degrees of Freedom (DoFs) space of possible hand ar-
ticulations. In our work, we decompose this space into six articulation subspaces
(6 DoFs for the palm and 4 DoFs for each finger). We also label each finger with a
Gaussian model that is propagated between successive image frames. As confirmed
by a number of experiments, this divide-and-conquer approach tracks hand articula-
tions more accurately than existing model-based approaches. At the same time, real
time performance is achieved without the need of GPGPU processing. Additional
experiments show that the proposed approach is preferable for supporting the ac-
curate manipulation of virtual objects in VR/AR scenarios.

97

Project
학술활동

국제학회

FINDING THE OPTIMAL DESIGN CREATIVITY IN
CAR DESIGN FEATURES AGAINST BRAND STYLES

International Conference on Design Creativity (ICDC)
S. An, D.A. Min, J. Lee

Automobile designers strive to create novel and distinctive designs without sacri-
ficing the car’s inner brand style. Both novelty and brand styles are known to be
paramount to realizing market success, but this balancing depends on the intuition of
designers and managers, which poses uncertainty for companies. Such uncertainty
inhibits the ability or willingness of designers and managers to make creative de-
signs as a barrier to potential creative designs. Therefore, we aim to support design-
ers and managers to achieve optimal creativity without violating brand identity by
providing the quantitative reasoning on brand styles. In this study, we decomposed
car designs into design features and identified the visual significance of design fea-
tures on maintaining brand styles. Results show that the particular design features
with higher visual significance must be kept for brand recognition while others have
greater design freedom for innovative designs.

Hand Contact between Remote Users
through Virtual Avatars

The 29th International Conference on Computer Animation
and Social Agents (CASA)
Jihye Oh, Youjin Lee, Yeonjoon Kim, Taeil Jin, Sukwon Lee, Sung-Hee Lee

We present an avatar animation technique for a telepresence system that allows for
the hand contact, especially handshaking, between remote users. The key idea is
that, while the avatar follows the remote user’s motion normally, it modifies the mo-
tion to create and maintain hand contact with the local user when the two users try
to engage hand contact. To this end, we develop the support vector machine (SVM)-
based classifiers to recognize the users’ intention for contact interaction, and online
motion generation method to create realistic image sequence of an avatar to realize
the continuous contact with the user. A user study has been conducted to verify the
effect of our method on the social telepresence.

98

Project
학술활동

국제학회

HoloStation: Augmented Visualization and Presentation

SIGGRAPH ASIA 2016 Symposium on Visualization
Minju Kim, Jungjin Lee, Wolfgang Stuerzlinger, Kwangyun Wohn

As much as stories need to be told, images need to be presented. Although visualiza-
tions are meant to be self-explanatory, often enhancing their expressive power by
incorporating a certain degree of interactivity, visualized images today often fail to
encourage the active engagement of the user/audience. In many cases, interactive
interventions by a human presenter have the potential to drastically improve the
engagement with visualization. Rather than just showing the content, the presenter
then enhances information delivery, e.g., by providing the context of the visuali-
zation. In this paper, we propose a novel concept called augmented presentation
in which the human presenter occupies the same physical space as the visualized
information, thereby presenting and interacting with the visualized images seam-
lessly. Depending on the level of engagement the presenter’s role may vary: from a
simple storyteller to an augmented presenter who may be regarded as a part of the
visualized entity. To further the development of the new idea of augmented pres-
entation, we have designed, implemented, and user-tested a visualization system
named HoloStation. The presenter is placed between two projection screens: the
front one is half-mirrored and the rear one is a conventional wall screen. The 3D
stereoscopic images are rendered to appear in-between, thereby creating a coherent
3D visualization space filled with digital information and the human presenter. We
have conducted a controlled experiment to investigate the subjective level of immer-
sion and engagement of the audience with HoloStation compared to the traditional
presentation. Our results suggest that our new form of augmented presentation has
a potential not only to enhance the quality of information presentation but also to
enrich the user experience on visualizations.

99

Project
학술활동

국제학회

MELODY EXTRACTION ON VOCAL SEGMENTS USING
MULTI-COLUMN DEEP NEURAL NETWORKS

The International Society for Music Information Retrieval
Sangeun Kum, Changheun Oh, Juhan Nam

Singing melody extraction is a task that tracks pitch contour of singing voice in
polyphonic music. While the majority of melody extraction algorithms are based on
computing a saliency function of pitch candidates or separating the melody source
from the mixture, data-driven approaches based on classification have been rarely
explored. In this paper, we present a classification-based approach for melody ex-
traction on vocal segments using multi-column deep neural networks. In the pro-
posed model, each of neural networks is trained to predict a pitch label of singing
voice from spectrogram, but their outputs have different pitch resolutions. The final
melody contour is inferred by combining the outputs of the networks and post-pro-
cessing it with a hidden Markov model. In order to take advantage of the data-driven
approach, we also augment training data by pitch-shifting the audio content and
modifying the pitch label accordingly. We use the RWC ataset and vocal tracks
of the MedleyDB dataset for training the model and evaluate it on the ADC 2004,
MIREX 2005 and MIR-1k datasets. Through several settings of experiments, we
show incremental improvements of the melody prediction. Lastly, we compare our
best result to those of previous state-of-the-arts.

100